Martin WaldburgerCristian Morariu

Peter RaczJürgen JähnertStefan WesnerBurkhard Stiller

Grids in a Mobile World:Akogrimo’s Network and Business

Views

April 2006

University of ZurichDepartment of Informatics (IFI)Binzmühlerstrasse 14, CH-8057 Zürich, Switzerland ifi

TE

CH

NIC

AL

RE

PO

RT

–N

o.20

06.0

5

M. Waldburger, C. Morariu, P. Racz, J. Jahnert, S. Wesner, B. Stiller:Technical Report No. 2006.05, April 2006Communication Systems GroupDepartment of Informatics (IFI)University of ZurichBinzmuhlerstrasse 14, CH-8050 Zurich, SwitzerlandURL: http://www.csg.unizh.ch/

1

Abstract — The use of wireless networking technolo-gies has emerged over recent years in many application domains. The area of grids determines a potentially huge application domain, since the typical centralized comput-ing centers require access from anywhere, e.g., from field engineers who are situated in a wireless network domain. Thus, the integration of suitable business views on mo-bile grids, of grid views on available technologies, and network views in a fully IP-based network domain de-termines the key challenge. The Akogrimo project’s ar-chitecture developed, is outlined and discussed in this paper and provides the major details required to offer a fully integrated and interoperable solution for those three views of concern.

Index Terms — Mobile Grid, Accounting, Charging, IP-based Grid Services, Multi-domain, Dynamic Virtual Organization

I. INTRODUCTION

Mobility has become a central aspect of life for people in business, education, and leisure. Related mobile 3G network infrastructures and user communities have surpassed corre-sponding Internet figures. Independent of this development, grid technology is evolving from a niche market – solely addressing the High Performance Computing (HPC) domain – toward a framework useable within a broad business con-text. However, while affecting largely identical complex applications, determined as user and provider domains, until now the grid community has been basically mobility-unaware. Therefore, the Akogrimo project (Access to Knowledge through the Grid in a Mobile World) [2] aims to radically advance the pervasiveness of grid computing across regions, basically by leveraging the large base of mo-bile users. In order to achieve this goal and in addition to embracing layers and technologies, which are supposed to make up the so-called Next Generation Grids (NGG) such as knowledge-related and semantics-driven web services, Akogrimo architects and prototypes a blueprint of an NGG. Akogrimo exploits and closely co-operates with evolving mobile Internet infrastructures based on IPv6 and integrates security as well as dynamicity aspects.

From a technical point of view, Akogrimo addresses in an integrated manner mobility, Quality-of-Service (QoS), Au-thentication, Authorization, Accounting, Auditing, and Charging (A4C), and security functionalities, which are pro-vided by corresponding network-related Akogrimo middle-ware systems. To pursue this goal, Akogrimo addresses at the same time the Grid and the Internet with an integrative architectural view. From a user’s point of view, Akogrimo will provide the technologies and concepts to establish a ‘virtual home', with nomadic and mobile environments for solving complex problems across network technology and provider domains.

In generalizing the core grid concept – namely the re-source-sharing concept – Akogrimo patterns these environ-ments as Mobile Dynamic Virtual Organizations (MDVO) [34]. The MDVO concept incorporates network-identity-based concepts of personalization, profiling, privacy, secu-rity, and trust. Finally, from the provider’s and, thus, busi-ness point of view, an Akogrimo world provides new busi-ness models and opportunities, making NGGs a reality and preparing them commercially viable for future use.

Thus, this overview paper will address the essential busi-ness model classifications for mobile grids as well as the key grid and networking technologies. Based on this description the development of the Akogrimo architecture handles the actor and role interrelations within mobile grids, addressing business, grid, and network views. This approach covers key characteristics as well as major functional components re-quired and integrates a middleware design process with a business-driven view point. To enable a complete spectrum of grid services, the respective network support integrates mobility, QoS support, a SIP-based signaling infrastructure, authentication, authorization, accounting, auditing, charging, context management, and service discovery. Thus, the key concern of a mobile grid, the offering of integrated services, has been mitigated.

The remainder of the paper is structured as follows. Sec-tion II discusses the key backgrounds. While Section III outlines the basic Akogrimo Architecture, sections IV, V, and VI determine its main business, grid, and network mid-dleware-related aspects. Section VII finally summarizes the work and draws preliminary conclusions.

Grids in a Mobile World: Akogrimo’s Network and Business Views

M. Waldburger1, C. Morariu1, P. Racz1, J. Jähnert2, S. Wesner2, B. Stiller1

1 Communication Systems Research Group, Department of Informatics (IFI), University of Zurich, CH-8057 Zürich, Switzerland, [morariu|racz|waldburger|stiller]@ifi.unizh.ch

2 HLRS and RUS, University of Stuttgart, Germany D-70569 Stuttgart, Germany, wesner@hlrs.de, jaehnert@rus.uni-stuttgart.de

2

II. BACKGROUND AND RELATED WORK

The areas of interest for an integrated view of grids in the wireless domain require the careful study of next-generation grids as such and their networking technology applied. This has to be completed by a study of the usefulness of grids in the wireless networking domain under business perspectives.

A. Next Generation Grids for Business

Grid principles are no longer only applied to the domain of High Performance Computing (HPC) but they are devel-oped towards a general purpose utility technology with the focus on service virtualization [14], [15]. Beside this techno-logical development to change the underlying infrastructure from proprietary protocols towards standardized communi-cation based on web services [1] technology there is a trend to virtualization, not only of low level resources such as computing power and storage, but also higher level re-sources such as knowledge [7].

Another important dimension is the development of busi-ness models for grid-based infrastructures. The European research project Akogrimo is leveraging its attempts to im-plement a commercially sustainable grid solution for tele-communications operators and service providers supporting mobile users and workers on these developments. Akogrimo, thus, envisages mobile knowledge grids for business – sub-sequently referenced to as mobile grids for reasons of simplicity. Knowledge in this context refers to pervasive access to knowledge, reflected by adaptive workflows, rather than artificial intelligence, such as reasoning components.

Following the definition of the key properties for Grids given by Foster in [13] Akogrimo defines a Mobile Grid as:

“A Mobile Grid consists out of resources that are not subject to centralized control, while it supports all kind of mobility (such as device, user, and session) and communi-cates using standard, open, general-purpose protocols and interfaces in order to deliver non-trivial and optimized Quality-of-Service (QoS) depending on the current context of the resource or the user. Furthermore, a Mobile Grid is communicating with the underlying network in both direc-tions enabling cross-layer cooperation between Grid mid-dleware and the network.”

In other words, both mobile and fixed grid systems are re-flected from an economic viewpoint by a virtual organiza-tion (VO). The term virtual organization was originally de-fined as an organizational model [31] in the domain of eco-nomics. Combining this with the idea of controlled resource sharing, as envisaged by the grid community [12] and the requirements from Business-to-Business (B2B) interactions [31] a more generic view is needed if mobility aspects should be considered. Mobile grids are characterized by two key functional extensions of fixed grid systems: First, the support of mobile grid nodes provides for pervasive, loca-tion-independent access to knowledge. Secondly, mobile grid nodes may dynamically join and leave the VO. Both aspects, mobility and dynamicity, show impacts on the one hand on respective business opportunities, e.g. by imple-menting context-sensitive adaptive workflows, on the other

hand they demand for enhancements of the VO concept, which does not reflect mobile and dynamic aspects as yet. This includes for example session models allowing for tem-porary interruption of services, when a mobile node is un-covered by a communications network. Accordingly, mobile grids are represented from a business point of view by Mo-bile Dynamic Virtual Organizations (MDVO), which are perceived as VOs that reflect mobile and dynamic aspects and that are defined in dependence on [22], [23], [33] as follows.

“An MDVO is a temporary or permanent coalition of geographically dispersed potentially mobile individuals, groups, organizational units, or entire organizations. They pool resources, capabilities, and information in order to contribute to the operational goals of the VO according to the dynamically established contractual agreements that are accountable and chargeable across administrative domains, incorporating mechanisms for parameterized and secure authentication as well as authorization.” However, note that the fully dynamic contractual negotiations required to oper-ate an MDVO in an open environment may be replaced by out-of-band, pre-negotiated contracts, which are pre-established between major organizations. Concerning mobil-ity support, the extension of mechanisms and algorithms required to maintain a VO covers mobility management, identity management, and mobility support protocols. Thus, a new instance of a VO, the MDVO forms the integrated concept for Akogrimo.

B. Networking Technology

The Akogrimo architecture sketched in section III is based on state of the art network technologies, covering (a) AAA (Authentication, Authorization, and Accounting) for access control and accounting, (b) Mobile Internet Protocol version 6 (MIPv6) [19], (c) SIP (Session Initiation Protocol) [30] for session handover and signaling, and (d) Differenti-ated Services (DiffServ) [4] as well as IntServ for QoS res-ervations. Each of those is briefly addressed below on the context of mobile grids.

(a) AAA systems provide standard means for user authen-tication, authorization of service access and accounting of service usage. These tasks are essential in commercial net-works and service provisioning environments [27]. The ge-neric AAA architecture [9] defines network components and their interactions on an abstract level, together with a pro-posal of a layered structure for AAA protocols. The RADIUS (Remote Authentication Dial-In User Service) protocol [29] was designed initially to support AAA func-tions in dial-up and terminal server access services, and it is the most widely deployed AAA protocol in current network infrastructures. Accounting support in RADIUS is specified in [28], which is restricted to network access services. Di-ameter [5], which is considered as the next generation AAA protocol, is based on RADIUS, however being more flexible and providing reliable data transfer, failover mechanisms, better error handling, and security mechanisms [5]. The Di-ameter protocol consists of the generic base protocol and various Diameter applications. The Diameter base protocol

3

defines Diameter entities and specifies common functional-ities, including Diameter message delivery, capability nego-tiation between Diameter nodes, and error handling. Diame-ter applications, such as NASREQ [6], enable the flexible extension of the protocol, defining service-specific com-mands and data units. Similar to RADIUS, Diameter mes-sage parameters and data units are coded in AVPs (Attrib-ute-Value-Pair) allowing the protocol to be extended in a flexible manner. Besides Diameter clients and servers, the protocol defines agents, which provide more flexible means in the message forwarding, supporting relay, proxy, redirect, or translation services. Additionally, Diameter enables, in contrast to RADIUS, server-initiated messages. Finally, Di-ameter supports secure communication based on IPSec (Internet Protocol Security) [20] and TLS (Transport Layer Security) [10] in intra- and inter-domain scenarios.

Security Assertion Markup Language (SAML) [25] is an XML-based framework for web services that enables the exchange of authentication and authorization information. A method for transporting SAML messages over Diameter protocol is proposed in [24].

(b) MIPv6 [19] defines a protocol that allows mobile nodes to be reachable while moving around in the IPv6 Internet. The new protocol simplifies the Mobile IPv4 archi-tecture by eliminating the need of a foreign agent (FA) and enabling the binding of a mobile’s node care-of address to its home address.

(c) SIP [30] defines an application layer control protocol for creating, managing, terminating sessions between two or more participants. The possible sessions include IP teleph-ony, multimedia content distribution, multimedia confer-ences, and instant messaging. SIP messages used to create sessions contain session descriptors that are used for nego-tiation of content media types. The SIP architecture is based on SIP proxies, which are elements concerned with routing messages to a user’s current location, authenticating and authorizing users for services, implementing session-routing policies and providing service-specific features to end-users. SIP implements a registration function which allows users to upload their current location for use by SIP proxies.

(d) Finally, DiffServ and IntServ define mechanisms for delivering QoS in access networks (IntServ) as well as in core networks (DiffServ). In [3], a framework is proposed for delivering end-to-end IntServ services over heterogene-ous DiffServ core networks.

C. Business Models for Current Grid Systems

Grid business models vary in their actual embodiment from instantiation to instantiation. However, grid business models consist – as business models do in general – of the following building components [32]:

• A description of involved actors and assigned business roles.

• An overview of various business flows, in particular product, information, and service flows.

• The value proposition, showing benefits for actors to be implemented.

• Sources of revenue, that indicate potentials for earn-ings, based on financial flows within a VO.

Besides these generic constituents of business models, specific characteristics for grid business models have to be considered. Due to the fact that grid systems offer non-tangible products, based on electronic services, grid eco-nomics are highly interrelated with electronic commerce business models. For e-business models, various taxonomies exist in parallel which is exemplarily documented by an em-bracing overview in [8]. These taxonomies are not directly comparable in the first place. However, the integrative ap-proach followed in [11], identifies 12 common criteria for classifying electronic business models. This allows depicting typical characteristics of today’s grid systems, in particular traditional high performance computing- and data-oriented grid computing as well as service grids that allow service virtualization in VOs. Accordingly, Table 1 determines a grid business model classification for grid computing and service grids, as representatives of current grid systems. As Table 1 outlines those mentioned classification criteria ap-plied on the existing HPC-oriented grid computing and VO-driven service grid approaches, the respective business model classification for next-generation mobile grids, thus reflecting the MDVO case drawn in section II.A, has been determined as follows:

Mobile grids will target B2C markets in a 1-to-Many way, in fact a Many-to-Many way, which however is hidden from service users, since virtualization is implemented by means of service aggregation. Services, thus, are composed over a distributed value chain, that is characterized by business process execution and knowledge-integration in hierarchies. Targeting mass markets requires a variety of charging mod-els and tariff schemes to be supported, since various differ-ent market segments with the need for customized pricing mechanisms are envisaged. Customization is however not limited to pricing, but it includes user and device context information so that workflows adapt to context changes, which in turn has consequences on the required security level of business flows: In an MDVO, where sensitive con-text information is exchanged across administrative do-mains, where high-level resources such as knowledge are integrated and made accessible to mobile users, security aspects in the single domains as well as with respect to in-formation flows between domains are of utmost importance – in particular for trust-building, privacy issues and com-mercial service provision supported by A4C mechanisms. Aiming at offering pervasive access to knowledge leads to challenging issues to be addressed as knowledge by defini-tion is embedded in a certain field of application where it is needed to solve a specific task. Whoever is able to provide knowledge according to context information, thus, qualifies for a differentiation strategy. Providers however offering basic (grid) services with a low level of integration only, are assumed to follow a cost-leadership strategy by focusing on core competencies. High-level resource coordination in an MDVO will involve various content elements, so that high data volumes are expected to be transmitted. If users then

4

are able to solve complex problems right on scene, clearly determines a high level of innovation in the mobile grid, which puts successfully operating grid service aggregators and telecommunication providers into a strong position.

Driven by those given indications on next-generation mo-bile grid business models, an overview of the Akogrimo overall architecture is given in section III, and detailed views subsequently are drawn for business, grid, and network mid-dleware aspects in sections IV, V, and VI respectively.

Table 1: Grid Business Model Classification for Existing Grid Systems (Criteria Based on [11])

CRITERIA DESCRIPTION

CR

ITE

RIA

TYPICAL VALUES

BUSINESS MODEL CHARACTERISTICS FOR EXISTING GRID SYSTEMS

Determines whether the client is related to an end user or a provider of a service which in turn is sold to other clients.

Use

r R

ole

End-user-related or resel-ler. B2B or Business-to-Consumer (B2C)

Clients usually are related to end us-ers, however not individuals but rather organizations, i.e. B2B customer rela-tions are maintained. A grid service provider for example might offer to solve data- or computational-intensive simulation tasks in automotive.

Denotes the number of involved service providers and service consumers

Inte

ract

ion

Patt

ern

1-to-1, 1-to-Many (one service provider, many service consumers), Many-to-1 (many service providers, one service consumer) or Many-to-Many [32].

1-to-1 or 1-to-Many in case of grid computing. By forming a virtual or-ganization, as in service grids, Many-to-1 and Many-to-Many become pos-sible, which however is not visible for service consumers. Due to today’s grids’ focus on niche solutions or non-commercial research projects, 1-to-1 and Many-to-1, respectively, are pre-vailing.

Describes the basic offer-ing characteristics.

Nat

ure

of O

ffer

ings

Content/information, service or complete prod-uct (including e.g. after-sales services)

Offerings usually embrace electronic services, such as computational power in grid computing. In a VO, various service and content providers are able to cooperate, whereas a service aggre-gator bundles basic services. Complete products typically are out of scope, since the offered electronic services represent only one step in the user’s value chain.

Specifies the financial terms, under which a user can access the service.

Pric

ing

Syst

em

Dynamic or fixed pricing with sub-categories such as free access, usage rate, subscription fee or auc-tions.

The pricing scheme highly depends on the actual organizational configuration of a grid-based solution as well on the type of offered services. Both, VO-externally and VO-internally, fixed pricing schemes, based on service usage, are regarded as significant choices under the assumption that relevant usage parameters are avail-able to be metered, accounted, and charged.

Determines how much a service is customized to a user’s individual require-ments.

Lev

el o

f C

usto

miz

atio

n

Full range from mass content to completely customized content.

Since current grid systems are charac-terized by 1-to-1 or Many-to-1 busi-ness relations, whereas B2B relations prevail and users are end-users (cf. respective classification criteria), mass markets are not targeted. Conse-quently, offered services are custom-ized to the user’s individual require-ments.

CRITERIA DESCRIPTION

CR

ITE

RIA

TYPICAL VALUES

BUSINESS MODEL CHARACTERISTICS FOR EXISTING GRID SYSTEMS

Depicts the organizational alignment with respect to service provision

Eco

nom

ic C

ontr

ol

Full range from self-organization to hierarchi-cal design.

Independent from the existence of a VO, the organizational structure is assumed to be hierarchical within business grids. A central component manages the distribution of tasks to be completed. From a user’s viewpoint, in a commercial environment, only one organization is assumed to main-tain direct customer relations, thus hiding organizational details of a potential VO from the user.

Identifies the needed security-level to monitor and verify purchases.

Lev

el o

f R

equi

red

Se-

curi

ty

Low or high

With an orientation towards direct B2B customer relations in a 1-to-1 manner, the required level of security for monitoring and verifying pur-chases is assessed to be low. In the case of service grids (VO-enabled, Many-to-1) however, monitoring has to be increased VO-internally among the various VO members.

Qualifies the degree of integrating and transform-ing components over different value chain steps.

Lev

el o

f V

alue

Int

egra

tion

Low or high

In a VO, several steps of the VO-internal value chain are integrated from the services provided by various VO members. The VO itself offers these electronic services in a bundled form to a user which in turn integrates them into its own value chain. Thus, assuming a service grid with service provisioning in a VO, the level of value integration is considered to be high. The opposite (i.e. a low level of value integration) however is the case in traditional, purely high performance computing-centric grid computing.

Reflects the envisaged competitive strategy.

Val

ue/C

ost O

ffer

ings

Differentiation (adding value) or cost leadership (low-cost or low-price strategy)

Selling purely computational power implements a cost leadership strategy where services are imitable and offer only little value added, they however are offered to customers at comparably inexpensive conditions. In service grids, a higher level of added value is reached by offering aggregated, appli-cation domain-specific services, which allows the grid service provider to implement a differentiation strategy.

Considers the expected volume of data traffic that is generated.

Scal

e of

T

raff

ic

Low or high

Grid systems are by design useful for solving computationally and data-intensive tasks. Accordingly, most grid services are expected to generate high data traffic rates.

Assesses whether existing business is directly trans-ferred to its electronic version or the electronic business offers new func-tionality [32].

Deg

ree

of I

nnov

a-ti

on

Low or high

Both forms of current grid systems, grid computing and service grids, offer a high level of innovation, since new functionality is gained compared to non-electronic business. This includes high performance computing facilities and virtualization of electronic ser-vices in a VO.

Answers whether buyer or seller is in a stronger position (e.g. with regard to negotiations).

Pow

er D

istr

ibut

ion

Buyer or seller

At this early stage of commercializa-tion, distribution of power is difficult to be assessed. However, commercial grid service providers, representing sellers, are not yet numerous which puts them into a better bargaining position, assuming a constant or grow-ing demand.

5

III. AKOGRIMO ARCHITECTURE

The key components of the Akogrimo Architecture are shown in Figure 1. The overall architecture is composed from four major building blocks, where three of them are in a layered structure and the fourth one has a vertical coordi-nation role. On the lower layer, the mobile IP network is depicted. The key components consist of the IPv6-based mobile network components, the network QoS provisioning components, as well as network management components. The key role of this layer it to provide logical interfaces to the higher layer in order to communicate context information to the Akogrimo architecture’s higher layers and to execute the service needs from the higher layer.

On the middleware layer service discovery, real-time ses-sion management, context management and authentication, authorization, accounting, and charging form key building blocks. This layer glues basically the two worlds, that of grid and that of a commercial mobile network in order to deploy grid services in a commercialized environment.

The grid infrastructure service layer consists of compo-nents that are generally seen as the fundamental elements of an OGSA-based grid infrastructure [12]. However, for Akogrimo these components have to be adapted to commu-nicate with components on the lower layers in an integrated manner and to provide that information from higher layers down to the network infrastructure to ensure an efficient interoperation. Furthermore, additional components needed for business grids such as the management of Service Level Agreements and policy management is placed in this layer.

The application support layer provides, as mentioned above, besides its function to provide generic application domain-independent supporting services for grid-enabled applications, also the necessary components enabling the management of the cross-layer cooperation. This functional-ity is covered by the VO management block that enables set-up, configuration, operation, and dissolution of operational VOs, all of which is supported by the workflow enactment block. Support of this VO lifecycle and the linked lifecycle of the dynamically established or pre-negotiated Service Level Agreements and contracts define the necessary actions on the lower layers. For example, the integration into an operational VO requests the necessary settings to enable access with appropriate rights to the resources of the part-ners in the VO. Moreover, the defined Service Level Agreements need to be enacted and managed on the lower

layers, which requires possibly the reservation of appropriate bandwidth on the (access) network layer and the configura-tion of AAA servers to react on validations of them.

IV. BUSINESS VIEW

From a business viewpoint, Akogrimo’s interest is set on economic aspects of commercial mobile grids, which are characterized by two key functional extensions of fixed grid systems: First, the support of mobile grid nodes provides for pervasive, location-independent access to knowledge. Sec-ondly, mobile grid nodes may dynamically join and leave the VO. Both aspects, mobility and dynamicity, show impacts on one hand on the respective business opportunities, e.g. by implementing context-sensitive adaptive workflows, on the other hand they demand for enhancements of the VO con-cept, which does not reflect mobile and dynamic aspects as yet, resulting in the respective definition of MDVOs (cf. Section II.A.).

Mobile grids are reflected by MDVOs from an organiza-tional viewpoint, but other perspectives also have to be con-sidered in order to gain an embracing view of relevant busi-ness aspects, particularly with regard to mobile grid business models. Following a top-down approach, starting at the or-ganizational alignment of mobile grids and going down to the underlying technical infrastructure, two structural parts besides the MDVO are identified, namely grid infrastructure and network infrastructure. These layers are drawn from an economic viewpoint and they are not to be perceived as di-rect equivalents to the respective architectural layers shown in Figure 1.

With respect to the first element of business models (cf. section II.C), i.e. involved actors and assigned roles, only MDVO and grid infrastructure are regarded to cope purely with grid business models. However, since these depend on communications infrastructure to be available, respective actors and roles on network infrastructure’s layer have to be included as well. Even though actors and role models show specific characteristics depending on the chosen scenario, generic constituents relevant for any mobile grid’s role model are assumed to be present in all mobile grid scenarios. This assumption bases on mobile grids’ distinctive charac-teristics from Akogrimo’s commercial perspective. This means that Akogrimo scenarios adopt a mobile grid vision where, first, presumably only end-users are mobile, which does not imply that user nodes are service consumers only, but computation- or data-intensive tasks are performed on fixed grid nodes. Second, in a commercial environment, single grid services are bundled by a service aggregator that maintains end-user relations, thus taking exclusively the role of a grid solution provider towards end-users. And thirdly, grid services are aggregated from a distributed supply chain that embraces n-tier grid service providers, whereas the role of a grid service provider can show more specific instantia-tions, such as the one of a content provider. Figure 2 depicts the organizational alignment being typical for Akogrimo mobile grid scenarios, incorporating actors and roles from an MDVO and grid services perspective.

OpVO Application Support Layer Grid Infrastructure Services Layer

Network Middleware Services Layer

Network Services Layer

BaseVO Optimization Framework Execution

Management

Service Quality and Agreement

Management

Data Management

Policy Management

Semantic based Service Discovery

Network Management

QoS Provision

IPv6/MIPv6 Infrastructure

SIP Session Manager A4C Context

Manager

OpVoBroker

SLA Definition

Workflow Enactment

VO Manager (membership)

Figure 1: The Akogrimo Architecture

6

Actors and roles from the network infrastructure layer

however are not explicitly shown in Figure 2 for reasons of simplicity, though assumed to be present in every scenario, since network service providers, represented for example by mobile network operators, play the indispensable role of a communications enabler. Accordingly, the following generic business roles are identified as the basic set being relevant for mobile grids that include Business-to-Consumer rela-tions:

• End-user (potentially a separate organizational entity than the end-customer)

• Grid Service Aggregator • Grid Service Provider • Network Service Provider Having denoted those four basic business roles, business

models for mobile grids are analyzed subsequently in further detail using the three remaining business model building blocks, which embrace business flows, the respective value propositions, as well as potential sources of revenue (as pre-sented in section II.C). Regarding value proposition, Table 2 lists the identified benefits for every basic business role, thus allowing to document potential values to be envisaged and achieved.

Table 2: Value Proposition for the Basic Set of Mobile Grid Business Roles

BU

SIN

ESS

RO

LE

VALUE PROPOSITION

End

-Use

r

End-users profit from being able to perform knowledge-intensive tasks while not being bound to a specific location anymore. Thus, services remain available even when changing locations and can be resumed after a temporary loss of network coverage. Services are reflected by context-aware, adaptive workflows, implementing various types of mobility, such as user, terminal, and service mobility [22]. Furthermore, end-users are not only able to consume service, but also to provide them, e.g., by first retrieving content relevant to perform a current task, working on the content, and delivering respective results back.

BU

SIN

ESS

RO

LE

VALUE PROPOSITION

Gri

d Se

rvic

e A

ggre

ga-

tor

The grid service aggregator is the producer and the vendor of a mobile grid solution. This solution makes value-added services available for end-users in form of a product. As a producer, atomic services are aggregated to service bundles. This implies that the producer runs and maintains the Akogrimo service delivery platform. As a vendor contractual relationships to-wards end-users, here in terms of customers, are maintained. Thus, the vendor qualifies for billing. Furthermore, the vendor copes with marketing activities and customer service.

Gri

d Se

rvic

e Pr

ovid

er

Grid service providers focus on their individual core competen-cies. Depending on the grid service characteristics, either both or only one, economies of scale and economies of scope, allow a grid service provider to implement a differentiation or cost leadership strategy while keeping transaction costs as low as possible. By the use of open, standards-based interfaces, grid service providers are able to provide their grid services in many VOs.

Net

wor

k Se

rvic

e Pr

ovid

er

Network service providers offer technology-transparent access to communications infrastructure, which forms the basis for grid services on top of communications services. Mobile net-work operators excel as both, network service providers and grid service aggregators, due to their experience in customer relations and marketing, allowing them to move from pure access providers to providers of complex services.

Business flows and possible resources of revenue are on one hand highly interrelated, whereas on the other hand, specific indications highly depend on the chosen scenario. In general, business flows are based on technical requirements to be met. This includes guarantees with respect to QoS or non-repudiation of messages exchanged among VO mem-bers. Single business processes, regardless whether executed within an administrative domain of a grid service provider or seen from an aggregated VO perspective, all result in busi-ness flows, which in turn cause financial claims that have to be settled.

Therefore, from a process angle, A4C (see Section VI.C is regarded as key element: Authentication, authorization, ac-counting, charging, and billing document for every provider the complete process from service instantiation to finally billing a provider’s customer. Hence, a grid accounting model, composed of meaningful accountable units that re-flect process costs and resource usage is needed. Account-able units are useful on the one hand with regard to service optimization and auditing, on the other hand for price-building and billing purposes, influencing finally direct sources of revenue. In a VO however, where organization members remain legally independent, business-critical ac-counting information is not assumed to leave administrative domains, so that only aggregated charging records are ex-changed.

Having presented various business-related aspects of commercial service provision in an MDVO leads to an in-depth investigation of the Akogrimo architecture, both from a grid and network middleware viewpoint as the two key elements by which business potentials are implemented. This includes for instance SLA management and workflow en-actment on grid level (cf. section V) and A4C as well as context management on network middleware layer (cf. sec-tion VI).

Figure 2: Generic Mobile Grid Actors and Roles in Akogrimo Scenarios

B2B B2C

Legend: End-user Grid Service

Grid Services Aggregator

Mobile Grid

Nodes

First Tier

Supplier

Second Tier

Supplier

N-Tier Suppliers

7

V. GRID VIEW

The deployment of grids has already achieved a signifi-cant number of deployments within organizations driven by the motivation to move from large mainframes to clusters aiming at a reduction of the Total Cost of Ownership (TCO). Also in purely academic environments, where the sharing of high performance rare resources is the driver for the grid solution, cross-organizational solutions are available. How-ever, grid solutions that enable the sharing of resources across organizational boundaries are not yet very common. The major obstacles preventing the wide take up are related to the complexity of user management, cross-organizational accounting, auditing and charging of used resources and security considerations.

A. Web Services Base Technology

All major grid toolkits nowadays are based on web ser-vice technology. Akogrimo is, despite its new challenges, coming from the business and network layer aiming to align with the major trend in grids and Business-to-Business rela-tions to rely on cross organizational businesses based on web services. Besides those base technologies as defined in the WS-I Profile Akogrimo can rely on the wide range of composable specifications including WS-Addressing [35], Web Services Base Notifications [26] and from the Web Service Resource Framework [17]. Besides other specifica-tions these are supported by all WSRF-based grid toolkits such as the Globus Toolkit 4 [16] or WSRF.NET [36]. Those have been selected as base technologies for the Akogrimo integration platform. One of the major design decisions that had to be taken within Akogrimo was on which of the layers (cf. Figure 1) the virtualization interface had to be designed. This interface provides the underlying resources and services through web service interfaces.

The originally anticipated, strictly layered approach to have this boundary between the Network Middleware Layer and the Grid Infrastructure Services Layer has been proofed to be not beneficial. Instead, and in line with the paradigm of Service Oriented Architectures (SOA) resources in the Net-work and Network Middleware Layer are exposed partially as Web Services and can be used by services from the Grid Infrastructure, the Grid Application Support “Layer”, and application-specific services.

B. State and Asynchronous Communication

A major differentiator between stateless, distributed ap-plications and Grid Applications rely on the WS-RF specifi-cations for exposing state information and WS Base Notifi-cation for a more peer-to-peer like communication pattern between grid nodes. This approach is based on topic-based subscription and asynchronous message delivery.

For example, the context of a current user device might be exposed using WS-RF allowing to query and manipulate the state and also to allow these states to be transferred in case the user changes his device and wants to transfer its state to the new device. Furthermore, if the context includes value such as “Current Bandwidth” or “Screen Resolution” other grid Nodes might want to subscribe to these state changes.

Within the example of context change propagation the gen-eral concept of Akogrimo of cross-layer co-operation can be demonstrated. The collected state information from the net-work layer (see section VI.D) using a wide range of different mechanisms is virtualized by a web service interface and is provided to context consumers using a standardized web service interface, independent on how this information had been collected. As a result standard workflow engines or any application service can subscribe to these notifications and adapt their behavior accordingly.

C. Service Level Agreements (SLA) and Execution Management Systems (EMS)

Another basic building block for moving from e-science grids to business grids is the support of service level guaran-tees to the consumer of services. Akogrimo is relying in this case on the results of the GRAAP working group of the Global Grid Forum namely WS-Agreement and WS-Agreement Negotiation [18]. As part of these specifications the possibility to negotiate the content of a Service Level Agreement (SLA) is described. In the volatile context of Akogrimo, e.g., where service providers might move from a high- to low bandwidth environment, the acceptance of SLAs can only be done based under certain conditions (e.g., the minimum bandwidth) and through active management of information technology and network resources. Within Akogrimo, besides the management of computational and data resources the execution management system (EMS), which received the SLA to be enacted, is also able to benefit from virtualization interfaces provided by the network layer, since it is enabled to select from different sets of QoS bun-dles. Thus, if a consumer demands a certain SLA this might cause a change in its access network quality class or poten-tially even in a change of the access network provider.

Akogrimo is approaching SLA and QoS clearly with a concept of bilateral agreements between consumers and pro-viders. While these agreements might be brokered they de-termine clearly electronic contracts between consumers and providers. This point is emphasized, since many approaches within the Grid domain aim for a global or VO-wide re-source optimization strategies, ignoring potentially cus-tomer-specific behavior of provider.

D. Workflows, Operational VOs, and Application Ser-vices

A common workflow language for web service-based or-chestration is the Business Process Execution Language (BPEL). As described above all grid nodes within Akogrimo are potentially subject to context changes and may conse-quently require a change in the execution of the workflow. For example, if the user changes from a powerful work-station that is capable of rendering complex medical data to a PDA the workflow needs to be adapted, e.g., by adding a rendering service that delivers a small picture of the ren-dered data together with a set of meta data optimized for the low resolution client. This might involve the discovery, con-figuration, and integration of a new member of the opera-tional virtual organization (see [23] for details) and a switch

8

of the current workflow. Furthermore, application services might be forced to change their behavior, e.g., through re-ducing the frequency of data updates or the provision of additional textual metadata. In general, the dynamic and timely adaptation of services on the application layer is a key issue for the success of a mobile grid, since they show a major impact on the user’s experience.

VI. NETWORK MIDDLEWARE VIEW

In the Akogrimo network, mobile users will be roaming across different access networks belonging to different ad-ministrative domains. Further, there is no permanent network coverage expected which leads to disconnected operations scenarios in order to maintain membership of an MDVO.

The underlying network and its components are required to provide the necessary infrastructure to deliver services to users as well as to provide supporting services for network and service providers. Figure 3 gives an overview of the Akogrimo network architecture. Users connect to the net-work via different end-devices, including portable devices and stationary terminals. Access networks provide wired or wireless connections via different physical access technolo-gies. The core network connects the access networks, pro-vides connectivity towards other networks and comprises network components that provide network and grid services. The grid middleware and the grid application infrastructure provide grid services. Figure 3 does not show network do-main boundaries but the depicted components support inter-domain communication and can be deployed in several do-mains. If multiple network providers are present, handovers among the different networks are supported, assuming that it is underlined by a commercial agreement.

Figure 3: Akogrimo Network Architecture

Selected parts of the Akogrimo network architecture are described in the following sections in detail.

A. Mobility and QoS support

One of the main challenges of Akogrimo is providing seamless access to grid services and applications in a mobile environment. The following three different types of mobility are identified and supported [22]:

• Device mobility: In a mobile communication environ-ment users are allowed to move throughout different networks having access to the same set of services.

• User mobility: Users of a mobile grid should have ac-cess to their services regardless of the device they use or the place and technology they use to access these services.

• Session mobility: Session mobility refers to seamless transfer of an ongoing service session from a device to another.

Device mobility is provided by mobile IPv6 extended with fast handover support [21]. The home agent (HA) in the core network serves as a location register for mobile termi-nals. User and session mobility are supported by the overall authentication process integrated in the network and the SIP infrastructure. Besides mobility the Akogrimo network infra-structure provides also QoS support. According to Figure 3 access routers (AR) in the access network provide the inter-face for end-terminals to connect to the network. The AR performs access control to network resources based on user authentication and meters network service usage. It receives QoS requests from the mobile terminal, controls QoS signal-ing and communicates with the QoS Broker. The QoS Bro-ker is the central entity to manage QoS within the network. It decides on QoS requests, manages available network re-sources and exchanges these information with other QoS Brokers. Mobility and resource management are integrated in order to support QoS management during handovers and to ensure that handovers between different access networks are possible.

B. SIP Infrastructure

The integrated SIP-based communication is provided by the SIP infrastructure, supporting a common session man-agement and session mobility. The central part of the infra-structure is the SIP server in the core network, which con-tains the SIP proxy and the SIP registrar for user registration (cf. Figure 3). Separate SIP proxies can be deployed in sev-eral access networks to distribute the load of the SIP server. The SIP server also provides presence information of the user to the context manager which collects and distributes these information to other network components.

C. A4C

In Akogrimo, the A4C system provides the necessary functionality for authenticating users, authorizing access to services, accounting and charging for service usage, based on agreed SLAs, as well as auditing. By its heterogeneous nature, in terms of services and service providers (SP) in-volved, a VO needs strong support from the A4C infrastruc-ture for managing inter-domain security and inter-domain accounting and charging. Emerging from a network technol-ogy, the A4C will deliver AAA services not only to network services, but also to grid-based services.

A mobile grid infrastructure aggregates services from multiple SPs in order to build new applications to be deliv-ered to end-users. The new applications will require QoS support across the different domains and will need to be

9

charged according to the quality experienced by the service customers. Such aggregations require trust agreements be-tween SPs (for simplicity network providers are also seen as SPs) at the A4C infrastructure level. Trust agreements are similar to the roaming agreements between mobile operators and allow a user to consume services from multiple SPs while keeping a business relationship with a single SP - his home domain. A user usually has an initial contact with his home domain represented by signing a contract between the two parties, and then a permanent security relationship be-tween them. The trust relationship between the user and his home domain and the trust relationships of that domain and other SPs enable the user to establish temporary, transitive trust relationships with new SPs thus allowing the user to access services provided by other SPs. The newly created trust relationship however does not imply any business rela-tionship between the user and the new SP. Challenges in designing and implementing the A4C system arise from the need to extend mechanisms developed ini-tially for network services in order to deliver their function-alities to mobile grid services. Akogrimo’s A4C infrastruc-ture is based on the Diameter protocol and it follows the IETF AAA generic architecture. The main components of A4C are the A4C server, the A4C client and the SAML au-thority. Their physical deployment in an existing network and grid infrastructure is depicted in Figure 4.

The A4C server performs authentication of users, au-thorization of network services, accounting of service usage and charging of service consumption. Besides these tasks, it also handles the inter-domain communication with A4C servers of other SPs based on existing cross-domain trust agreements.

Figure 4: A4C View of the Network Architecture

The SAML authority is a component built based on [25] and is used by the A4C server for providing single-sign-on support across multiple domains. Its task is to generate and check SAML assertions during user authentication sessions.

The A4C client is a module used by any network or grid component that requires A4C functionalities. Its main task is to establish and manage communication with the A4C server in order to provide access to the A4C infrastructure to any component that needs it.

The following paragraphs shortly describe the tasks per-formed by the A4C system.

1) Authentication: Two different authentication mechanisms are used in

Akogrimo. Network authentication is the first one that is used when a user attaches to an access network (AN). Dur-

ing the network authentication a user is identified based on a username and a password. If the AN belongs to the home SP the authentication is processed by the local A4C server, oth-erwise the local A4C server will redirect the request to the home domain of the user for further processing. As result of a successful authentication the user receives a SAML artifact (IDToken) which will be used for further service request as a proof of being an authenticated user. An IDToken is gener-ated by the SAML authority from the user’s home domain during the authentication process, and it can only be vali-dated by the SAML authority that generated it. The valida-tion of IDTokens is the second authentication mechanism used in Akogrimo.

2) Authorization In Akogrimo, service authorization has two distinct as-

pects: on the one hand there is network services authoriza-tion which allows the user to have network connectivity and on the other hand there is a more complex authorization mechanism that is used inside a VO for enabling the interop-eration of different services belonging to different SPs. The A4C server will perform network services authorization, while the authorizations needed for the VO internal opera-tions will be performed by VO management components. The Akogrimo components that will make use of the au-thorization mechanisms of the A4C server are the QoS bro-ker and the VO manager. Whenever a user attaches to an access network, the QoS broker responsible for that access network will request from the A4C details on the QoS that needs to be de1livered to the user. The VO manager will contact the A4C server whenever a user will request to join a VO. The VO manager requests the generic user profile (GUP) of a user. Based on a user profile, further authoriza-tion decisions will be made at the VO level.

3) Accounting The A4C server controls and manages the accounting

process. It collects and stores accounting records, which describe the usage data for a certain service session. An ac-counting session is related to the service usage of a user. Service usage data is gathered on different components by metering network traffic, grid resources utilization or other user activities. The components which perform metering and generate accounting data are access router (AR) and the grid monitoring component. A commercial mobile grid platform for being successful is required to offer different levels of QoS and charge its users based on the QoS they receive dur-ing service provisioning. The A4C system performs this task by specifying for every service metrics that define the qual-ity of that service as well as the degree of the service usage. The acceptable values for these metrics are part of an SLA agreement and are continuously monitored and accounted for during the service session.

4) Auditing Services in Akogrimo are monitored for the parameters

negotiated in the SLA. The auditing component gathers in-formation on the SLA violations occurring during service consumption. The data recorded during auditing processes will be used later on during the charging process.

10

5) Charging Based on the accounting records collected during account-

ing process and the charging specification stored in the A4C server, the charging subsystem calculates the charge for each service session and creates associated charging records. A charging record contains the result of the charge calculation and it includes attributes like the username, accessed service, domain where the service was provided, currency used, and the charge for the service. To support service provisioning in a multi-domain environment where users receive a single bill from their home provider, the exchange of charging records between providers is required. Charging records are used both for the transfer of end user charges and for the ex-change of settlement data between providers.

D. Context Management

The Akogrimo context manager gathers basic context data, processes and refines it according to common Akogrimo context requirements and allows context con-sumer to add extension modules to handle domain-specific context inference. The overall Akogrimo context manage-ment approach is shown in Figure 5.

The context manager gathers raw context data from dif-ferent context data sources across well-defined interfaces. Given the variety of possible context data sources, the con-text manager needs to handle multiple protocols for collect-ing data from different sources. The context manager has a set of rules (policies) describing the logic for context gather-ing (e.g., which sources provide context, at what schedule, action to take when something fails). These rules are config-ured through an administrative interface. The context man-ager refines raw context data to provide standard formats (e.g., geographical coordinates, temperatures), performs processing according to common Akogrimo requirements and stores the context data in a context database.

E x te n s io nm o d u le s

C o n te x t g a th e r in g

Q u e r ie s S u b s c r ib e /n o t i f y

B a s ic c o n te x t d a t a s o u r c e s

P r o c e s s in g a n d s to r a g e

C o n te x t g a th e r in g

Q u e r ie s S u b s c r ib e /n o t i f y

B a s ic c o n te x t d a t a s o u r c e s

P r o c e s s in g a n d s to r a g eD o m a ins p e c i f i cc o n te x ts o u r c e s

C o n te x tc o n s u m e r s

Figure 5: Context Management Functional Layers

Context consumers can either execute queries towards the context manager to get specific context information or sub-scribe for notifications about changes in certain context data. The context manager is responsible for distributing context data to context consumers accordingly. To handle domain-

specific context inference, the context manager offers inter-ested parties interfacing with extension modules. One exam-ple is the translation of location coordinates to a map of a building, such that the context will also specify which room a person is in and what physical facilities he is able to use. Context extension modules also allow context consumers to combine basic context information with additional domain-specific context sources, e.g. data from specific medical sen-sors.

E. Service Discovery

The functional role of the service discovery service (SDS) is to allow a service requestor to find an appropriate service provided by a service provider. A service provider must de-scribe the service capabilities (service description, SD) and publish the service through the service discovery service. The SDS maintains a list of available services and their ca-pabilities. Since within the Akogrimo architecture very di-verse services must be handled, a unified service description is a key issue. The different types of service discovery pro-tocols target different types of services, such as people, places, or web services. In order to embrace the greatest number of SD types several protocols will be used, which may encounter service directory protocols and their access protocols.

Within the Akogrimo environment, the general service discovery system (GSDS) is in charge of supplying a par-ticular service from a particular client request. GSDSes have to deal with services of very different nature. On the one hand, with the so-called local services/resources such as printers, beamers, virtual hard disks, and on the other hand with the grid services that can be defined as stateful web services, which are suited for the semantic grid.

The system in charge of discovering local resources (ser-vices) is called local service discovery system (LSDS), whereas grid service discovery system (GrSDS) is the name for the module dealing with proper Grid Services as pre-sented in Figure 6.

Service Providers ( Proximity services )

Service Requestor

General Service Discovery System ( GSDS ) ( Agent / broker )

Local Service Discovery System ( LSDS )

SLP UPnP Tag Badge

Services Device People Places , Things ...

Grid Service Discovery System ( GrSDS ) - >

UDDI ++

Service Providers ( WS )

Web Services ,

Grid Services

Publish Services

Figure 6: Service Discovery

11

VII. SUMMARY AND CONCLUSION

In this overview paper, a new approach on grid business models – mainly driven by the Akogrimo context, however, taking general requirements into account – has been devel-oped in which the detailed list of classification criteria as well as key characteristics have been exploited. Based on an extensive related work discussion, the Akogrimo Architec-ture has been outlined by addressing business, grid, and net-work middleware views. Driven by a three layer model, the network services layer, the network middleware services layer, and the grid infrastructure services layer, the grid ap-plication support in terms of virtual organization (VO) memberships, workflow enhancements, and Service Level Agreements (SLA) can be granted. This approach results in a highly flexible system, which addresses modern mobility requirements at the same time.

To show the integrated understanding of networking and grid aspects, three different views have been introduced. While the business view of Akogrimo focuses on economic aspects of commercial mobile grids as such, the grid view determines the services, infrastructure, and technology avail-able for now. Finally, the network view provides insights into the supportive distributed systems, which provide mo-bility and Quality-of-Service (QoS) support, signaling sup-port, and authentication, authorization, accounting, auditing, and charging (A4C) support, context management support, and service discovery support. Thus, these elements form the base for a single technology solution, in which networking and grid approaches are harmonized in the sense of a com-mon A4C system and joint signaling system.

As the project Akogrimo is being implemented in its cur-rent state, the architectural design steps undertaken have resulted in an integrated view of grids and networks. This integration is believed to leverage off the effort seen so far in pure grid approaches, which do not address explicitly the potential knowledge of underlying networking technologies. Neither the support of roaming users nor the integrated pro-visioning of full-fledged and automated accounting schemes have been visible so far in existing systems. Thus, Akogrimo provides a unique solution for accessing knowledge through the mobile link.

In the near future there will be a prototype available, which addresses key areas of the grid and networking inte-gration. Driven by implementation restrictions and at the same time standardization requirements, a SOAP and diame-ter-based system will form the basis for experimental work and evaluations.

Acknowledgements

This work has been performed in the framework of the EU IST project Akogrimo “Access to Knowledge through the Grid in a Mobile World” (FP6-2004-IST-2-004293). The Akogrimo partners are acknowledged highly for their collaboration.

REFERENCES

[1] D. X. Adamopoulos, C. A. Papandreou: Enabling Web Services: Towards Service Grids. Ninth International Symposium on Com-puters and Communications 2004 (ISCC 2004), Vol. 1, June & July 2004, pp 38-43.

[2] Akogrimo: Access to Knowledge through the Grid in a Mobile World. European Union project FP6-2004-IST-2-004293, http://www.mobilegrids.org, October 2005.

[3] Y. Bernet, P. Ford, R. Yavatkar, F. Baker, L. Zhang, M. Speer, R. Braden, B. Davie, J. Wroclawski, E. Felstaine: A Framework for In-tegrated Services Operation over Diffserv Networks. IETF RFC 2998, November 2000.

[4] S. Blake, D. Black, M. Carlson, E. Davies, Z. Wang, W. Weiss: An Architecture for Differentiated Services. IETF RFC 2475, December 1998.

[5] P. Calhoun, J. Loughney, E. Guttman, G. Zorn, J. Arkko: Diameter Base Protocol. IETF RFC 3588, September 2003.

[6] P. Calhoun, G. Zorn, D. Spence, D. Mitton: Diameter Network Ac-cess Server Application. IETF RFC 4005, August 2005.

[7] M. Cannataro, D. Talia: Semantics and Knowledge Grid: Building the Next-generation Grid. IEEE Intelligent Systems, Vol. 19, No. 1, January-February 2004, pp 56-63.

[8] R. Cao, Y. Chai: The Study of the Electronic Commerce Business Models. 2004 International Conference on Systems, Man and Cyber-netics (SMC 2004), Vol. 5, October 2004, pp 4089-4094.

[9] C. de Laat, G. Gross, L. Gommans: Generic AAA Architecture. IETF RFC 2903, August 2000.

[10] T. Dierks, C. Allen: The TLS Protocol Version 1.0. IETF RFC 2246, January 1999.

[11] M. Dubosson-Torbay, A. Osterwalder, Y. Pigneur: E-Business Model Design, Classification, and Measurements. Thunderbird International Business Review, Vol. 44, No. 1, January-February 2002, pp 5-23.

[12] I. Foster: The Anatomy of the Grid: Enabling Scalable Virtual Or-ganizations. First IEEE/ACM International Symposium on Cluster Computing and the Grid, May 2001, pp 6-7.

[13] I. Foster: What is the Grid? A Three Point Checklist. GridToday, Vol. 1, No. 6, June 2002, http://wwwfp.mcs.anl.gov/~foster/Articles/WhatIsTheGrid.pdf.

[14] I. Foster, C. Kesselman: The GRID: Blueprint for a New Computing Infrastructure. Morgan Kaufmann Publishers Inc., 1999.

[15] I. Foster, C. Kesselman, J. M. Nick, S. Tuecke: The Physiology of the Grid. Grid Computing. Making the Global Infrastructure a Reality, John Wiley & Sons, 2003, pp 217-249.

[16] The Globus Alliance: The Globus Toolkit. http://www.globus.org/toolkit/, March 2006.

[17] The Globus Alliance: The WS-Resource Framework. http://www.globus.org/wsrf/specs/ws-wsrf.pdf, May 2004.

[18] Global Grid Forum (GGF): GGF-GRAAP Working Group Docu-ments. http://forge.gridforum.org/docman2/ViewCategory.php?group_id=71&category_id=659, March 2006.

[19] D. Johnson, C. Perkins, J. Arkko: Mobility Support in IPv6. IETF RFC 3775, June 2004.

[20] S. Kent, R. Atkinson: Security Architecture for the Internet Protocol. IETF RFC 2401, November 1998.

[21] R. Koodli: Fast Handovers for Mobile IPv6, IETF RFC 4068, http://www.ietf.org/rfc/rfc4068.txt, July 2005.

[22] A. Litke, D. Skoutas, T. Varvrigou: Mobile Grid Computing: Changes and Challenges of Resource Management in a Mobile Grid Environment. 5th International Conference on Practical Aspects of Knowledge Management (PAKM 2004), December 2004.

[23] C. Loos, S. Wesner, J. Jähnert: Specific Challenges of Mobile Dy-namic Virtual Organizations. eAdoption and The Knowledge Econ-omy: Issues, Applications, Case Studies, 2005.

[24] A. Olivereau, A.F. Gomez Skarmeta, R. Marin Lopez, B. Weyl, P. Brandao, P. Mishra, H. Ziemek, C. Hauser: An Advanced Authoriza-tion Framework for IP-based B3G Systems. 14th IST Mobile & Wireless Communications Summit, Dresden, June 2005.

[25] Organization for the Advancement of Structured Information Stan-dards (OASIS): Security Assertion Markup Language, SAML, http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=security, October 2005.

12

[26] Organization for the Advancement of Structured Information Stan-dards (OASIS): Web Services Base Notification 1.2 (WS-BaseNotification). Working Draft 03, http://docs.oasis-open.org/wsn/2004/06/wsn-WS-BaseNotification-1.2-draft-03.pdf, June 2004.

[27] C. Rensing, H. Hasan, M. Karsten, B. Stiller: AAA: A Survey and a Policy-based Architecture and Framework. IEEE Network Maga-zine, Vol. 16, No. 6, November 2002, pp 22-27.

[28] C. Rigney: RADIUS Accounting. IETF RFC 2866, June 2000. [29] C. Rigney, S. Willens, A. Rubens, W. Simpson: Remote Authentica-

tion Dial In User Service (RADIUS). IETF RFC 2865, June 2000. [30] J. Rosenberg, H. Schulzrinne, G. Camarillo, A. Johnston, J. Peterson,

R. Sparks, M. Handley, E. Schooler: SIP: Session Initiation Protocol. IETF RFC 3261, June 2002.

[31] L. Schubert, S. Wesner, T. Dimitrakos: Secure and Dynamic Virtual Organizations for Business. eAdoption and The Knowledge Econ-omy: Issues, Applications, Case Studies, 2005.

[32] P. Timmers: Business Models for Electronic Markets. Journal on Electronic Markets, Vol. 8, No. 2, April 1998, pp 3-8.

[33] M. Waldburger, B. Stiller: Toward the Mobile Grid: Service Provi-sioning in a Mobile Dynamic Virtual Organization. IFI Technical Report 2005.07, ftp://ftp.ifi.unizh.ch/pub/techreports/TR-2005/ifi-2005.07.pdf, August 2005.

[34] S. Wesner: Towards and Architecture for the Mobile Grid. it-information Technology, Vol. 47, No. 6, June 2005, pp 336-342.

[35] World Wide Web Consortium (W3C): Web Services Addressing (WS-Addressing). http://www.w3.org/Submission/ws-addressing/, August 2004.

[36] WSRF.NET Forum: WSRF.NET. http://www.cs.virginia.edu/~gsw2c/wsrf.net.html, March 2006.